LLM-enhanced by choice
LLM use is planned as an explicit opt-in enhancement.
The default command must not call OpenAI, Anthropic, OpenRouter, Copilot, Ollama, or any other model provider.
When LLM mode is introduced, it should require explicit flags such as:
branchbrief --llm --provider openaiLLM output should improve narrative fields only. Deterministic facts such as commits, changed files, diff stats, and risk signals remain the source of truth.